Prediction of particle-laden pipe flows using deep neural network models

Haghshenas, Armin; Hedayatpour, Shiva; Groll, Rodion

doi:10.1063/5.0160128

Cited by 3 publications

(1 citation statement)

References 65 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Schwarz et al 104 used a deep neural network (DNN) model for capturing erosion from the resulting particle trajectories and speeds (i.e., rebound model) for particleladen flow through turbines. Similarly, Haghshenas et al 105 used DNN models for particle-laden flows in sediment transport lines.…”

Section: ■ Proposed New Pathway For Scale-upmentioning

confidence: 99%

Fluidized Bed Scale-Up for Sustainability Challenges. 2. New Pathway

Chew,

Cocco

2024

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

Despite more than 100 years of commercialization of wide-ranging fluidized bed reactors, scale-up tools and methods have remained quite similar. To exploit the benefits of fluidized beds for the time-critical sustainability challenges, scale-up has to be implemented quicker and better. Correspondingly, a companion Part 1 (Ind. Eng. Chem. Res. 2024, 63, 2519−2533 reviewed the evolution of the tools used in scaling up fluidized beds. Leveraging that, the current Part 2 aims to first overview the traditional pathway for scale-up and then propose a new pathway. Notably, instead of the traditional way of focusing on a linear sequence of progressively larger units, the emphasis is on the Phases of Discovery, Research, and Development, which apply the new tools consistently, as well as address risk mitigation and economics throughout. Based on an acrylonitrile case study, a Monte Carlo analysis indicates the new proposed pathway offers more promising economic feasibility, with net present value over 20 years (NPV 20 ) of $310MM higher, along with 35% and 42% reductions in start-up and break-even times, respectively. The increase in costs for incorporating modeling efforts is insignificant compared to the overall benefits with respect to time and economics.

show abstract

Section: ■ Proposed New Pathway For Scale-upmentioning

confidence: 99%

Fluidized Bed Scale-Up for Sustainability Challenges. 2. New Pathway

Chew,

Cocco

2024

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

show abstract

Characterizing and predicting bluff-body solid fuel ramjet performances via shape design and multi-objective optimization model

Zhang,

Zhao,

Shi

et al. 2023

Physics of Fluids

View full text Add to dashboard Cite

In this work, we propose a rapid optimization approach to examine its application potential for the design and performance prediction and optimization of a solid fuel ramjet (SFRJ) with a bluff body. For this, the shape of the bluff body is parameterized first using the non-uniform rational B-spline method. We then develop a model for predicting SFRJ performances by incorporating both levy motion-gradient descent and support vector regression methods. It is found that a faster prediction is achievable, while the average error is maintained to be less than 5%. We then develop a multi-objective optimization model by considering the full thrust and minimum total pressure loss (TPL). The optimization model is examined using the non-dominated sorting genetic algorithm. A cost parameter is also created to facilitate the tradeoffs between the thrust and TPL in the Pareto front, when different bluff-body design configurations are considered. The present results reveal that an increase in the cost parameter will elevate the turbulence intensity within the SFRJs while drawing the incoming air closer to the fuel surface, resulting in an increase in thrust and regression rate, but the TPL will also increase. When prioritizing the TPL reduction in the design stage, the optimized solution reduces TPL by 50%. Meanwhile, the net thrust is shown to be decreased by less than 3.5%. Furthermore, flow-field investigation reveals that the improved performance of the optimized SFRJ is due to more uniform flow velocity gradients around the bluff body and a reduced rear vortex, resulting in reduced momentum loss. Our proposed optimization approach's robustness has been further confirmed with consistent performances, as the ramjet inlet speed varies over a broad range. It shows that our approach has great potential to be applied for the SFRJ performance prediction and optimization, being operated under various conditions.

show abstract

Self-supervised learning based on Transformer for flow reconstruction and prediction

Xu,

Zhou,

Bian

2024

Physics of Fluids

View full text Add to dashboard Cite

Machine learning has great potential for efficient reconstruction and prediction of flow fields. However, existing datasets may have highly diversified labels for different flow scenarios, which are not applicable for training a model. To this end, we make a first attempt to apply the self-supervised learning (SSL) technique to fluid dynamics, which disregards data labels for pre-training the model. The SSL technique embraces a large amount of data (8000 snapshots) at Reynolds numbers of Re = 200, 300, 400, and 500 without discriminating between them, which improves the generalization of the model. The Transformer model is pre-trained via a specially designed pretext task, where it reconstructs the complete flow fields after randomly masking 20% data points in each snapshot. For the downstream task of flow reconstruction, the pre-trained model is fine-tuned separately with 256 snapshots for each Reynolds number. The fine-tuned models accurately reconstruct the complete flow fields based on less than 5% random data points within a limited window even for Re = 250 and 600, whose data were not seen in the pre-trained phase. For the other downstream task of flow prediction, the pre-training model is fine-tuned separately with 128 consecutive snapshot pairs for each corresponding Reynolds number. The fine-tuned models then correctly predict the evolution of the flow fields over many periods of cycles. We compare all results generated by models trained via SSL and models trained via supervised learning, where the former has unequivocally superior performance. We expect that the methodology presented here will have wider applications in fluid mechanics.

show abstract

Prediction of particle-laden pipe flows using deep neural network models

Cited by 3 publications

References 65 publications

Fluidized Bed Scale-Up for Sustainability Challenges. 2. New Pathway

Fluidized Bed Scale-Up for Sustainability Challenges. 2. New Pathway

Characterizing and predicting bluff-body solid fuel ramjet performances via shape design and multi-objective optimization model

Self-supervised learning based on Transformer for flow reconstruction and prediction

Contact Info

Product

Resources

About